Sky Cover: Shining Light on a
Gloomy Problem
Jordan Gerth
Cooperative Institute for Meteorological Satellite
Studies, Space Science and Engineering Center
University of Wisconsin – Madison
23 July 2013
1
The Ultimate Problem
Not a clear definition
Poor instrumentation for observing high cloud
Observers occasionally fail to correct it
Meteorologists often neglect sky cover due
challenges with other parts of the forecast
• Numerical weather prediction models are grid
based and cloud grids are often too bimodal
• Satellites see the tops of clouds, but not
necessarily the surface sky cover
•
•
•
•
2
Defining Sky Cover
The National Weather Service (NWS) web site
defines “sky cover” as “the expected amount of
opaque clouds (in percent) covering the sky
valid for the indicated hour.”
• No probabilistic component.
• No definition of “opaque cloud” or “cloud”.
• The implication is cloud coverage of the celestial
dome (all sky visible from a point observer).
3
Forecasting Sky Cover
The NWS’ National Digital Forecast Database
(NDFD) contains the gridded operational
forecast for sky cover. Issues with the national
one-hour forecast include:
– Clear areas with non-zero cloud cover
– Vastly different cloud classifications for similar
cloud scenes
– Lack of spatial continuity between forecast areas
– Temporal trends do not match observations
4
5
6
Observing Sky Cover
• Geostationary satellites are helpful in
assessing sky cover because of good
– Spatial coverage (4 km)
– Temporal continuity (15 minutes)
• Effective cloud amount (or effective cloud
emissivity) is a close proxy to sky cover, with
some exceptions
• Satellites observe cloud top-down (high cloud
first), humans see bottom-up (low cloud first).
7
The GOES Imager Effective Cloud Amount (ECA) is the standard effective cloud emissivity
product from the GOES Imagers, valid at the indicated time.
8
The GOES Imager Celestial Dome ECA is an average of the standard effective cloud emissivity
within a box of 11 by 11 pixels, centered on each grid point, valid at the indicated time.
9
The GOES Imager Sky Cover Product is a time-average of the celestial dome ECA within a onehour window. The valid time begins at the time indicated on the plot. The average is all scans
after the valid time, within one hour.
10
Surface Observations
• These are all sky
observations taken from
surface stations.
• The fractional cloud cover
from each observation is
converted from the
reported cloud type in the
METAR.
• Multiple observations
within a one-hour
window following the
valid time, and/or within
10 km, are averaged.
Reported Coverage
Assigned Value
Clear
0%
Scattered
40%
Broken
75%
Overcast
100%
Obscured
100%
Thin Scattered
25%
Thin Broken
60%
Thin Overcast
90%
KMKE 230052Z 17007KT 10SM FEW050
SCT095 BKN250 26/20 A2973 RMK
AO2 SLP062 T02560200
11
12
13
Blended Sky Cover Analysis
For a given point:
• If satellite sky cover product or surface
observation analysis indicates a clear sky, the
blended sky cover analysis value is clear.
• If the surface observation analysis is higher than
the satellite sky cover product, the surface
observation analysis value is used.
• Otherwise, the two measurements are blended
with a weight depending on the distance to the
nearest non-clear surface observation.
14
15
Optimal Sky Cover Analysis
• Uses a linear solver to minimize the mean
absolute difference between the blended sky
cover analysis and the NDFD one-hour
forecast, subject to some constraints, for a
selected set of points.
• The main constraint is that the piecewise
linear function must be continuous, and that
the minimum (0) and maximum (100) values
must be anchored.
16
The GOES Imager Optimal Sky Cover is a linear optimization of the sky cover product, intended
to minimize absolute error when compared the National Digital Forecast Database (NDFD)
one-hour forecast. The valid time and range coincides with the sky cover product.
17
18
Optimizing Sky Cover
• Model output can be optimized to produce
the desired quantity.
• In this case, the objective is to decrease the
mean absolute difference between the
optimal sky cover analysis and a linear sum of
model variables subject to coefficients and
scalars applied to non-zero mixing ratios.
• The coefficients are further constrained by a
tolerance constraint on the mean value.
19
Optimizing Sky Cover
Variables used:
– Relative Humidity (all
levels)
– Cloud Water Mixing
Ratio, Rain Water Mixing
Ratio, Snow Mixing Ratio
(all levels)
– Absolute Vorticity (200
hPa only), partitioned
into positive and
negative components
• Pressure levels used:
–
–
–
–
–
–
–
–
–
200 hPa
300 hPa
500 hPa
700 hPa
800 hPa
850 hPa
900 hPa
950 hPa
1000 hPa
20
21
The High-Resolution Rapid Refresh (HRRR) Total Cloud Cover is the model output cloud cover
from the analysis of the run initiated at the valid time.
22
23
24
Conclusions
• Current effort is focusing on producing three-, six-, and
nine-hour forecasts of sky cover from the HRRR using
the output from the optimization model
• Linear optimization may yield fruit in the investigation
of other problems in the atmospheric sciences where:
– Multiple optimized quantities are required
– A relationship between the optimized quantities is
understood
– Quantities are subject to constraints relative to each other
or supported by the science
• Questions? Comments? Jordan.Gerth@noaa.gov
25